[jira] [Commented] (BOOKKEEPER-673) Ledger length can be inaccurate in failure case

["Sijie Guo (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/BOOKKEEPER-673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13746596#comment-13746596 ] 

Sijie Guo commented on BOOKKEEPER-673:
--------------------------------------

is there any application relying on this length field rather than the LAC field? if this length field is a critical part as LAC, should we also fix it in 4.2.2?

from technical perspective, it should be an easy and straightforward fix: since we keep pending add ops in a queue, we could decrement the length when error out pending add ops during closing.
                
> Ledger length can be inaccurate in failure case
> -----------------------------------------------
>
>                 Key: BOOKKEEPER-673
>                 URL: https://issues.apache.org/jira/browse/BOOKKEEPER-673
>             Project: Bookkeeper
>          Issue Type: Bug
>            Reporter: Ivan Kelly
>             Fix For: 4.3.0
>
>
> Ledger length can be inconsistent if a ledger is closed by a writing client that encounters an error. For example, in a cluster with 3 bookies, and an ledger with a q3e3 configuration, if a bookie dies, the client will close the ledger when it fails to write an entry. However, it has already added the length of the failed entry to the local ledger length, and this is what is stored to zk.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira